KaigeRay RLlib: PPO+Action-Mask+Customized ModelsThis article shows how to integrate action mask and customized models to PPO in Ray RLlib. It contains the following steps:1d ago1d ago
KaigeRay RLlib: Action-Mask+DQNThis article, how to add discrete action mask into Ray RLlib DQN algorithm. We use Pytorch framework. It contains the following steps3d ago3d ago
KaigeNot Another Entry-Level Tutorial of Reinforcement LearningThis is a series of posts on ‘advanced bits’ of practical reinforcement learning. Topics will be covered:May 23May 23
KaigePyTorch Version of Life-Time Value PredictionA DEEP PROBABILISTIC MODEL FOR CUSTOMER LIFETIME VALUE PREDICTIONApr 27Apr 27
KaigePaper Reading: Imitation Learning with Concurrent Actions in 3D GamesThis paper is from SEED Team Electronic Arts. In this article, we put quotations from this paper to be a summaryApr 18Apr 18
KaigePaper Reading: Proto-Value-NetworksProto-Value Networks: Scaling Representation Learning with Auxiliary TasksApr 13Apr 13
KaigeHow to learn successor representations from data?This article summaries the methods in the literature o how to learn successor representations from dataApr 1Apr 1
KaigePaper Reading: Generalization successor features to continuous domains for multi-task learningAs the title shows, this paper demonstrate how to use the learned successor feature in continuous multi-task learning. I put some key…Mar 27Mar 27
KaigePaper Reading: Learning Successor States and Goal-Dependent Values: A mathematical ViewpointThis paper is a deep analysis on successor representation (SR). I put some key insights and points from this paper.Mar 26Mar 26